A keyword search method based on knowledge graph association search recommendation

Abstract

The application discloses a keyword retrieval method based on knowledge graph association search recommendation. The method can perform association search on keywords based on a knowledge graph, recommend corresponding associated object instances to users according to selected features of the users, and quickly recommend useful information to the users. The method comprises the following steps: A, receiving a keyword; B, searching for object instances corresponding to the keyword, and searching for multiple associated object instances of the object instances according to attribute similarity and / or relationship closeness; C, combining the object instances and the associated object instances into a knowledge graph; D, performing clustering analysis on object instance nodes in the knowledge graph according to time features, node type features, node attribute features and relationship features, and setting feature options for the knowledge graph according to the clustering analysis; E, outputting the knowledge graph and the feature options to a user; and F, selecting object instance nodes meeting selected features of the user from the knowledge graph and recommending the object instance nodes to the user.

Description

Technical Field

[0001] This invention relates to the field of keyword retrieval technology, and in particular to a keyword retrieval method based on knowledge graph-based association search recommendation. Background Technology

[0002] Knowledge graphs are semantic networks that reveal relationships between entities. Currently, some engineers in the power industry are using knowledge graph technology to construct knowledge graphs for the power business domain, enabling the visualization of various power equipment and their relationships. However, the power business involves various stages such as generation, transmission, transformation, distribution, and consumption, and the number of power equipment involved is usually enormous, with complex relationships. Treating each device as an object instance node results in a vast number of object instance nodes in the constructed knowledge graph, with intricate relationships, making it difficult for users to quickly extract useful information. Summary of the Invention

[0003] The technical problem this invention aims to solve is how to quickly recommend useful information to users and reduce the effort users spend on information acquisition.

[0004] To address the aforementioned technical problems, this invention provides a keyword retrieval method based on knowledge graph-based association search and recommendation, comprising the following steps:

[0005] Step A: Receive keywords input by the user;

[0006] Step B: Retrieve the object instance corresponding to the keyword from the knowledge graph database, and obtain multiple related object instances of the object instance from the knowledge graph database based on attribute similarity and / or relationship tightness;

[0007] Step C: Combine the object instances corresponding to the above keywords and their associated object instances into a knowledge graph;

[0008] Step D: Perform cluster analysis on the object instance nodes in the knowledge graph according to time features, node type features, node attribute features, and relationship features respectively. Set feature options for the knowledge graph based on the cluster analysis results. These feature options include one or more of the following: time feature options, node type feature options, node attribute feature options, and relationship feature options.

[0009] Step E: Output the knowledge graph and its feature options to the user;

[0010] Step F: Based on the features selected by the user, select object instance nodes that match the features from the graph and recommend them to the user.

[0011] Furthermore, in step F, object instance nodes that match the user's selected features are recommended to the user in real time.

[0012] Furthermore, step F specifically involves highlighting object instance nodes that match the aforementioned characteristics in the knowledge graph to recommend them to the user.

[0013] Furthermore, it also includes the following steps:

[0014] Step G: Based on the object instance node selected by the user, display multiple different types of association extension options for the user to extend the association of the selected object instance node;

[0015] Step H: Receive the association extension option selected by the user. This option carries the object instance node information selected by the user. Based on attribute similarity and / or relationship tightness, search for the associated object instances of the object instance node from the knowledge graph database, and recommend the object instances whose instance type belongs to the association extension option type to the user.

[0016] Furthermore, in step H, the object instance is specifically displayed as a node overlaid in the knowledge graph of the keyword.

[0017] Furthermore, the various types of association extension options specifically include five types of association extension options: entity, event, document, concept, and all.

[0018] The keyword retrieval method based on knowledge graph association search and recommendation described above can perform association search on keywords based on knowledge graphs. According to the features selected by the user, it recommends related object instances that match the features, thereby quickly recommending useful information to the user and reducing the effort that the user spends on information acquisition. Attached Figure Description

[0019] Figure 1 This is a flowchart of the keyword retrieval method based on knowledge graph association search recommendation provided by the present invention.

[0020] Figure 2 This is a schematic diagram of the retrieval process of the keyword retrieval method based on knowledge graph association search recommendation provided by the present invention. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to specific embodiments.

[0022] like Figure 1 The keyword retrieval method based on knowledge graph-based association search recommendation shown includes the following steps:

[0023] Step A: Receive keywords input by the user;

[0024] Step B: Retrieve the object instance corresponding to the keyword from the knowledge graph database, and obtain multiple related object instances of the object instance from the knowledge graph database based on attribute similarity and / or relationship tightness;

[0025] Step C: Combine the object instances corresponding to the above keywords and their associated object instances into a knowledge graph;

[0026] Step D: Perform cluster analysis on the object instance nodes in the knowledge graph according to time features, node type features, node attribute features, and relationship features respectively. Set feature options for the knowledge graph based on the cluster analysis results. These feature options include one or more of the following: time feature options, node type feature options, node attribute feature options, and relationship feature options.

[0027] Step E: Output the knowledge graph and its feature options to the user;

[0028] Step F: Based on the features selected by the user, select object instance nodes that match the features from the graph and recommend them to the user;

[0029] Step G: Based on the object instance node selected by the user, display multiple different types of association extension options for the user to extend the association of the selected object instance node;

[0030] Step H: Receive the association extension option selected by the user. This option carries the object instance node information selected by the user. Based on attribute similarity and / or relationship tightness, search for the associated object instances of the object instance node from the knowledge graph database, and recommend the object instances whose instance type belongs to the association extension option type to the user.

[0031] Furthermore, in step F, object instance nodes that match the user's selected features are highlighted in real time and recommended to the user.

[0032] Furthermore, in step H, the object instance is specifically displayed as a node overlaid in the knowledge graph of the keyword.

[0033] The keyword retrieval method based on knowledge graph association search recommendation described above is applied to the graph search and analysis platform. The following section uses the user's retrieval process on the graph search and analysis platform as an example to illustrate the specific steps of this method.

[0034] like Figure 2As shown, the user logs into the knowledge graph search and analysis platform's search interface on the client, enters keywords, and clicks search. Assuming the user enters the keyword "Substation No. 2," after receiving the search request, the knowledge graph search and analysis platform first finds the object instance corresponding to "Substation No. 2" from the knowledge graph database. Then, it further searches the knowledge graph database for multiple related object instances with an attribute similarity higher than 80% and a relationship density higher than 80%. In a less preferred embodiment, other embodiments may search for related object instances with both attribute similarity and relationship density higher than 80%.

[0035] After retrieving the object instances and associated object instances for the keywords, the graph search and analysis platform combines these object instances and associated object instances into a knowledge graph. Then, it performs cluster analysis on the object instance nodes in the knowledge graph according to time features, node type features, node attribute features, and relationship features. Based on the cluster analysis results, it generates timeline statistical bar charts, node type statistical bar charts, node attribute statistical bar charts, and relationship type statistical bar charts. A feature selection operation area is also set up next to the knowledge graph (see...). Figure 2 These statistical bar charts are placed in the feature selection area, providing users with feature options. The graph search and analysis platform generates a knowledge graph with this feature selection area and outputs it to the user. Users can select corresponding features from time-axis statistical bar charts, node type statistical bar charts, node attribute statistical bar charts, or relationship type statistical bar charts for search recommendations. For example, if a user needs to know which electrical equipment was added to "Substation No. 2" in 2022, they can select 2022 from the time-axis statistical bar chart generated by the equipment addition time. The graph search and analysis platform then selects object instance nodes added in 2022 from the knowledge graph based on the time feature "2022," and highlights these object instance nodes in real time, thus recommending these nodes to the user.

[0036] Similarly, users can select desired node type features in the node type statistics bar chart, desired node attribute features in the node attribute statistics bar chart, or desired relationship type features in the relationship type statistics bar chart. The graph search and analysis platform will then select object instance nodes from the knowledge graph that match the user's selected features and recommend them to the user. Furthermore, users can select multiple features simultaneously, such as time features and node type features. The graph search and analysis platform will then recommend relevant object instance nodes to the user based on these time and node type features.

[0037] If a user wants to expand and explore the associations of a specific object instance node in the knowledge graph, they select that node. The graph search and analysis platform will then display five association expansion options: entity, event, document, concept, and all. For example, if a user wants to expand and explore related object instances of the event type, they would select the event type association expansion option. The graph search and analysis platform will then search the knowledge graph database for related object instances of the selected node, displaying the event-type related object instances as overlaid nodes in the knowledge graph, and recommending these related object instances to the user. This allows the user to perform further association searches based on keyword search results and obtain more useful information.

[0038] Similarly, if a user wants to discover related object instance nodes of entity type, document type, or concept type, they should select the corresponding type of association expansion option. If a user wants to discover related object instance nodes of all types, namely entity, event, document, and concept object instance nodes, they should select the all-type association expansion option. The graph search and analysis platform will then search the knowledge graph database for all related object instances of the selected object instance node, and display these related object instances as overlaid nodes in the knowledge graph, thereby recommending these related object instances to the user.

[0039] The keyword retrieval method based on knowledge graph association search and recommendation described above can perform association search on keywords based on knowledge graphs. According to the features selected by the user, it recommends related object instances that match the features, thereby quickly recommending useful information to the user and reducing the effort that the user spends on information acquisition.

[0040] The above description is merely an embodiment of the present invention and does not limit the scope of patent protection. Any non-substantial changes or substitutions made by those skilled in the art based on the present invention will still fall within the scope of patent protection.

Claims

1. A keyword retrieval method based on knowledge graph-based association search and recommendation, characterized by: Includes the following steps: Step A: Receive keywords input by the user; Step B: Retrieve the object instance corresponding to the keyword from the knowledge graph database, and obtain multiple related object instances of the object instance from the knowledge graph database based on attribute similarity and / or relationship tightness; Step C: Combine the object instances corresponding to the above keywords and their associated object instances into a knowledge graph; Step D: Perform cluster analysis on the object instance nodes in the knowledge graph according to time features, node type features, node attribute features, and relationship features respectively. Set feature options for the knowledge graph based on the cluster analysis results. These feature options include one or more of the following: time feature options, node type feature options, node attribute feature options, and relationship feature options. Step E: Output the knowledge graph and its feature options to the user; Step F: Based on the features selected by the user, select object instance nodes that match the features from the knowledge graph and recommend them to the user; Step G: Based on the object instance node selected by the user, display multiple different types of association extension options for the user to extend the association of the selected object instance node; Step H: Receive the user's selected association expansion option, which carries the information of the object instance node selected by the user. Based on attribute similarity and / or relationship tightness, search for related object instances of the object instance node from the knowledge graph database, and recommend related object instances whose instance type belongs to the association expansion option type to the user. In step H, specifically, the related object instances are displayed as nodes overlaid in the knowledge graph of the keyword.

2. The keyword retrieval method based on knowledge graph-based association search recommendation as described in claim 1, characterized in that: Step F specifically involves recommending object instance nodes that match the user's selected features in real time.

3. The keyword retrieval method based on knowledge graph-based association search recommendation as described in claim 1, characterized in that: Step F specifically involves highlighting object instance nodes that match the aforementioned characteristics in the knowledge graph to recommend them to the user.

4. The keyword retrieval method based on knowledge graph-based association search recommendation as described in claim 1, characterized in that: The various types of association extension options are specifically five types of association extension options: entity, event, document, concept, and all.

Images